Computational Social Science

I find computational social science [1] to be a nice term coined for the relatively new interdisciplinary field that can be summarized as computational methods applied on large datasets to investigate social sciences. Hence, it involves several subdisciplines: computational sociology, computational economics, computational linguistics, and even computational sociolingustics, culturomics and many others.

After attending the Third International Conference on Computational Social Science, IC2S2 2017, in Cologne, Germany, during the past week, I feel I can say some things about the current situation in the field. However, my blogpost will only be able to scratch the surface of some of the topics that resonated with me, hence, not being representative of the field. Given around 120 accepted talks and 80 posters, it was impossible to follow all the results. Nevertheless, I hope that my summary that follows showcases couple developments and trends that are worthwhile and interesting.

Social Media for Health

Ingmar Weber and Yelena Mejova (Qatar Computing Research Institute) summarized in their tutorial talk how researchers in the field investigated health.

One of the most severe diseases of today — depression, has been tackled a lot. In particular, to infer the likelihood of suicidal ideation, researchers used data from semi-anonymous support communities on Reddit [2]. They showed it possible to predict (to a certain degree) from previous discussions when someone will start having suicidal discussions. In another study, Instagram photos posted by depressed individuals were found more likely to be bluer, grayer, and darker. Moreover, people performed worse in predicting depression from Instagram photos, compared to the algorithms. In both cases, we see that algorithms running on online data could be detecting psychological diseases: how will be this used in the near future?

If discussed examples inspire you to develop interventions and support, then following examples show how interventions can be risky if not carefully designed. Namely, on Flickr, there exist unfortunate groups that support and promote anorexia (pro-anorexia).  Seemingly a good sign, researchers also found counter-groups (pro-recovery), that try to reach the members of the first groups and inform them of the negative consequences and dangers of starving oneself. However, research results show that such pro-recovery groups are only counterproductive, at the moment. They entrench the pro-anorexic individuals in their stance [4]. As another example, we heard of a short-lived project that aimed to warn the friends of individuals vulnerable to depression (using similar methods as described above). However, what happened is that some malicious people used this service to intentionally harm such vulnerable people. These examples show that no matter how good intentions you have, you need to carefully attend to their possible effects online — the effects can be unpredictable. Hence, for those of us who still want to help others, the (research) question is how to best design positive health interventions using social media?

Investigating Psychology using online data

Have you ever wondered how many of the reviews on Amazon or TripAdvisor are fake? Those same reviews that you might be basing your decisions on. The answer (for which I do not have a citation) is up to 30%. Nevertheless, I believe that the crowdsourced content platforms are still working well for my purpose —  likely because of the efforts by companies to deal with the fake contributions.

Now, one possible way to detecting fake reviews is using network analysis methods (such reviews have different patterns and frequency compared to real ones). However, during the IC2S2, I have learned about another method that is equally fascinating. Namely, there are established theoretical principles about deceptive statements versus true ones:

  • honest statements are richer in detail (The theory of Reality Monitoring),
  • they contain more contextual references to people, times and places (Criteria-based Content Analysis),
  • fake statements avoid information that have potential to be checked (Verifiability Approach).

Computational Social Science approach now is ‘just’ to develop methods that will evaluate reviews based on these three principles, and you have a fake reviews detection approach [5].

The study about psychological and personality profiles of political extremists [6] that fascinated me is the last one I will discuss herein.

You have probably, too, wondered like me — why some people hold as extreme views. While in some areas extreme views can potentially be useful or at least benign, in most of the areas, they are known to be harmful: either for the person holding the views, or for the people surrounding her, or both.

Recruitment into radical Islamic movements has renewed global interest in political extremist views. This time, given two competing psychological theories about profiles of political extremists, researchers used computational methods on large datasets to asses which theory agrees better with the data.

Nicely summarized to competing hypotheses are:

  • extremists differ psychologically from mainstream activists regardless of their left or right ideology (Collective Behavior Hypothesis),
  • left- and right-ideology activists differ psychologically from each other, independently on whether they are extremist or mainstream (Moral Foundations Hypothesis).

Perhaps surprisingly, researchers have found that the first hypothesis agrees with the (Twitter) data. If confirmed on other datasets, this result would mean, for instance, that radical pro-environmentalists or anarchists have more in common with Neo-Nazis or Neo-Confederates than one would perhaps expect.

While probably still too early to interpret in the above manner, I want to point to one last detail that I found incredibly curious in discussed study. Being different on all the Big Five Personality Traits (openness, agreeableness, consciousness, extroversion, neuroticism), political extremists are higher from all other types of people the researchers investigated on openness to experience. Given that openness is defined to include active imagination (fantasy), aesthetic sensitivity, attentiveness to inner feelings, preference for variety, and intellectual curiosity, how perplexed are you by the result? There are also moderate positive relationships of openness with creativity, intelligence and knowledge. However, another attribute of openness, related to psychological traits of absorption and hypnotic susceptibility, might seem more expected.


Hope that presented ideas have left you inspired and interested to read more from computational social science as they did with me.

[1] Lazer, David, Alex Sandy Pentland, Lada Adamic, Sinan Aral, Albert Laszlo Barabasi, Devon Brewer, Nicholas Christakis et al. “Life in the network: the coming age of computational social science.” Science (New York, NY) 323, no. 5915 (2009): 721.

[2] De Choudhury, Munmun, Emre Kiciman, Mark Dredze, Glen Coppersmith, and Mrinal Kumar. “Discovering shifts to suicidal ideation from mental health content in social media.” In Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems, pp. 2098-2110. ACM, 2016.

[3] Reece, Andrew G., and Christopher M. Danforth. “Instagram photos reveal predictive markers of depression.” arXiv preprint arXiv:1608.03282 (2016).

[4] Yom-Tov, Elad, Luis Fernandez-Luque, Ingmar Weber, and Steven P. Crain. “Pro-anorexia and pro-recovery photo sharing: a tale of two warring tribes.” Journal of medical Internet research 14, no. 6 (2012).

[5] Kleinberg, Bennett, Maximilian Mozes, and Arnoud Arntz. “Preprint: What’s in a name? Using named entities for verbal deception detection.” (2017).

[6] Alizadeh, Meysam, Ingmar Weber, Claudio Cioffi-Revilla, Santo Fortunato, and Michael Macy. “Psychological and Personality Profiles of Political Extremists.” arXiv preprint arXiv:1704.00119 (2017).

Leave a Reply

Your email address will not be published. Required fields are marked *